added ability to allow execution dag to be submitted and validated st… #96
Conversation
| execution_dag_schema = json.load(f) | ||
|
|
||
|
|
||
| def validate_schema(query): |
There was a problem hiding this comment.
Rename function to: prevalidate or similar -- something to indicate this is just a cursory validation.
Doc string along the lines of:
This performs a cursory validation to see if we have the basic format of an execution DAG. Note that this does not do any in-depth checks to see if [conforms to functions we have, security/injection, correct parameters, etc.], but just [what it does do]. [Prevent basic errors]
| query = dag | ||
| if not learning_observer.communication_protocol.schema.validate_schema(query): | ||
| debug_log(f'Submitted execution query is not a valid form') | ||
| funcs.append(DAGIncorrectFormat()) |
There was a problem hiding this comment.
I would consider including the validation exception in the DAGIncorrectFormat, so we can provide some useful debugging info to the client.
| return True | ||
| except jsonschema.ValidationError as e: | ||
| print(e) | ||
| return False |
There was a problem hiding this comment.
We might want to bubble the specific error back to the client in some way. I might consider continuing to raise an exception to be handled downstream. Perhaps, in pseudocode, something like: raise DAGIncorrectFormat(ValidationError.string_format())
| debug_log(f'Execution DAG, {dag_name}, not found.') | ||
| funcs.append(DAGNotFound(dag_name)) | ||
| dag = client_query.get('execution_dag', query_name) | ||
| if learning_observer.settings.settings.get('dangerously_allow_insecure_dags', False) and type(dag) == dict: |
There was a problem hiding this comment.
if type is string:
do stringy_stuff
if type is dict:
if not dangerously_allow_dags:
raise security_exception
do_dicty_stuff
else:
bad type
Or possibly:
def do_stringy_stuff()
pass
def do_dicty_stuff()
pass
dispatch = {dict: do_dicty_stuff, str: do_stringy_stuff}
if type(dag) not in dispatch: raise exception
dispatch[type(dag)]
* Working styling * LOStudentTable * Sample essays can now return essays of multiple types * PEP8 * Minor bug fix handling INF * Sync between devices * Working styling * LOStudentTable * Sample essays can now return essays of multiple types * PEP8 * Minor bug fix handling INF * Sync between devices * added js2py and fixed css file reference --------- Co-authored-by: Piotr Mitros <piotr@mitros.org>
Allow execution dags to include 1 level of dependencies. For example, the GPT dashboard now uses a dag within its module that references the Writing Observer dag.
…g_observer into comm-protocol-schema
|
Github is doing the weird pull request thing again. I will close this and reopen. |
…ructure